Actividades formativas

5.4.3.1 REO Duration, Property Value, and Built Environment

Table 7 and 8 shows the analytical results of the Cox proportional hazard model. In addition to the hazard ratio, the hazard coefficient was also included to aid the

interpretation of the results. Table 7 presents the results for all samples. Consistent with the literature (Pfeiffer & Molina, 2013), the coefficients on square footage, the number of bedrooms, and year built are significant. An increase in the square footage of a building by 100 and the number of bedrooms decreased the likelihood of selling REOs by 0.3% and 2.17% respectively. An increase in the year of construction increased the likelihood of an REO being sold by 0.2%.

Among the socio-economic attributes of neighborhoods, the race/ethnicity (Hispanic, black, and Asian) was found to be significant. When the population of

Hispanics and blacks in neighborhoods increased by 10 percentage points, the likelihood of REO properties being sold was lowered by 1.1% and 3.0%, respectively. Conversely,

127

the likelihood of an REO being sold increased by 2.04% when the population of Asians in neighborhoods increased by 10 percentage points. These results are consistent with the literature, showing that the racial/ethnic composition of neighborhoods significantly influences the length of time REO properties remain on the market (Y. Li & Walter, 2013; Pfeiffer & Molina, 2013). However, Y. S. Lee and Immergluck (2012) found that minority communities experienced faster sales. They noted that lower-valued REOs were more likely to be sold, and that such properties were generally located in minority communities. While the previous study (Y. S. Lee & Immergluck, 2012) found that properties in neighborhoods with a higher median household income increased the likelihood of being sold, this study did not find any statistical significance for median household income. Somewhat unexpectedly, a percentage increase of vacant units in a neighborhood increase the likelihood of an REO being sold by 3.77%. A possible explanation is that neighborhood vacancies may relate to negative home equity, and the decrease in home values might be positively associated with the likelihood of selling REOs.

Consistent with the literature, the results show that REO properties with a higher property value were less likely to be sold. The estimated hazard ratio of the property value variable was 0.9973, meaning that when a property value increases by $10,000, an REO property has a 0.27% lower likelihood of being sold, when holding other factors constant. Previous research indicated that investors were more likely to purchase low- value properties because of their high absorption rate in the market (Immergluck, 2012; Immergluck & Law, 2014).

128

For residential density, land-use mix, and street connectivity, the estimated hazard ratios were found to be significant almost exclusively in the upper percentile category. The single terms of those three domains were estimated as negative, indicating that REOs in more compact, mixed and accessible neighborhoods were less likely to be sold. However, the interaction terms with property value were estimated as positive, indicating that the effects of compact, mixed, and accessible neighborhoods on REO duration increase with property value. Compared to the middle level of residential density, REOs in the upper level of residential density had a 4.2% lower likelihood of being sold at the mean of the samples; however, the likelihood of being sold increased by 0.17 of a percentage point with the increase in property value. The likelihood of being sold became positive when a property value reached nearly $640,000.24 REOs in the upper level of mixed land-use areas had a 1.13% lower likelihood of being sold at the mean of the samples.25 The estimated interaction effect of land-use mix with property value was not statistically significant. Upper-level street connectivity was also estimated as negative in the single term and positive in the interaction term with property value. Upper-level street connectivity increased the hazard ratio by 0.12 of a percentage point with every $10,000 increase in property value.

For other built environmental attributes, the results found that the likelihood of REOs being sold was 2.38% greater in neighborhoods with bike lanes. Neighborhood

24 _{Based on the hazard coefficients, the property value (PV) of $640,000 was derived from the following} calculation: PV = 0.1082/0.0017. The coefficient of the upper percentile of residential density was - 0.1082, and the coefficient of the interaction was 0.0017.

129

safety influenced the likelihood of an REO being sold; however, crash-related safety had a significant hazard ratio of 0.60, meaning that REOs located in areas with higher crash rates were 40% less likely to be sold when the crash rate increased by one crash per acre. The estimated coefficient of park availability was positive but not significant.

Table 7. Results of REO Duration and Built Environments for All REOs

Variables Coeff. (S.E.) Haz. Ratio (S.E.) Z P-value 2009 -0.1728 (0.0098) 0.8414 (0.0083) -17.70 0.000 2010 -0.2407 (0.0107) 0.7861 (0.0084) -22.61 0.000 2011 -0.1150 (0.0121) 0.8915 (0.0108) -9.54 0.000 2012 0.0456 (0.0173) 1.0467 (0.0181) 2.65 0.008 2013 0.3516 (0.0475) 1.4213 (0.0675) 7.41 0.000 Sqft (/100) -0.0031 (0.0010) 0.9970 (0.0010) -3.34 0.001 Beds -0.0220 (0.0057) 0.9783 (0.0056) -3.90 0.000 Year built 0.0020 (0.0003) 1.0020 (0.0003) 8.47 0.000 Loan-to-value 0.0001 (0.0001) 1.0001 (0.0001) 0.66 0.512 Median income Middle -0.0152 (0.0122) 0.9850 (0.0121) -1.24 0.214 High -0.0176 (0.0175) 0.9826 (0.0172) -1.01 0.313 Hispanic (/10) -0.0080 (0.0032) 0.9921 (0.0032) -2.48 0.013 Black (/10) -0.0140 (0.0031) 0.9862 (0.0030) -4.59 0.000 Asian (/10) 0.0202 (0.0041) 1.0204 (0.0042) 4.95 0.000 Pop18 (/10) 0.0128 (0.0112) 1.0129 (0.0114) 1.14 0.252 Pop65 (/10) -0.0145 (0.0138) 0.9857 (0.0136) -1.05 0.293 Ownership (/10) 0.0068 (0.0036) 1.0068 (0.0037) 1.88 0.060 Unemployment (/10) 0.0118 (0.0103) 1.0119 (0.0104) 1.15 0.249 Vacancy (/10) 0.0370 (0.0100) 1.0377 (0.0104) 3.72 0.000 Mortgage (/10) 0.0163 (0.0161) 1.0164 (0.0163) 1.01 0.311 Active living (/10) -0.0146 (0.0175) 0.9856 (0.0173) -0.83 0.405 Crime density -0.0017 (0.0069) 0.9984 (0.0069) -0.24 0.809 Crash density -0.5043 (0.0880) 0.6040 (0.0532) -5.73 0.000

130

Table 7. Continued

Variables Coeff. (S.E.) Haz. Ratio (S.E.) Z P-value Property Value (PV) (/ $10,000) -0.0029 (0.0004) 0.9973 (0.0004) -7.17 0.000 Residential density Lower level 0.0086 (0.0200) 1.0086 (0.0202) 0.43 0.670 Upper level -0.1082 (0.0243) 0.8976 (0.0218) -4.47 0.000 PV × Residential density Lower level -0.0005 (0.0005) 0.9996 (0.0005) -1.07 0.286 Upper level 0.0017 (0.0006) 1.0017 (0.0006) 2.76 0.006 Land-use mix Lower level -0.0162 (0.0172) 0.9840 (0.0169) -0.94 0.346 Upper level -0.0421 (0.0192) 0.9588 (0.0184) -2.20 0.028 PV × Land-use mix Lower level 0.0002 (0.0004) 1.0002 (0.0004) 0.44 0.657 Upper level 0.0008 (0.0005) 1.0008 (0.0005) 1.62 0.106 Street connectivity Lower level -0.0655 (0.0205) 0.9367 (0.0192) -3.21 0.001 Upper level -0.0403 (0.0226) 0.9606 (0.0217) -1.79 0.074 PV × Street connectivity Lower level 0.0017 (0.0005) 1.0017 (0.0005) 3.63 0.000 Upper level 0.0012 (0.0006) 1.0012 (0.0006) 2.15 0.032 Bike lane availability 0.0235 (0.0080) 1.0238 (0.0082) 2.94 0.003 Park availability 0.0002 (0.0082) 1.0002 (0.0082) 0.02 0.981

Notes: N=73837, LR-Chi2=1704.87 (p<0.0001), Log-likelihood=-735576.78; The units of Property value, Sqft, Hispanic, Black, Asian, Pop18, Pop65, Unemployment, Vacancy, Mortgage, and Active living were adjusted to obtain valid coefficients and hazard ratios; Bold texts represent the statistical significance at the 0.05 level.

131

Figure 11 presents the plots of the estimated log hazards for the assessed property value, given the different percentile levels of built environmental attributes. The lines for the residential density and street connectivity showed similarities in trends. With the increase in property value, the estimated log hazards decreased in the low and middle levels of residential density and street connectivity. For the upper level, the slope of the lines was rare, and above certain property values, the estimated log hazards were higher than in the lower and middle levels. The plots for the upper level of residential density and street connectivity indicate that the likelihood of being sold increased with property values for REO properties located in compact and accessible neighborhoods. Land-use mix did not show any significant differences in the levels of built environments.

132

Residential Density Land-Use Mix Street Connectivity

Figure 11. Estimated Log Hazard Plots by Different Levels of Built Environmental Attributes

Residential Density Land-Use Mix Street Connectivity

133

5.4.3.2 REO Duration and Built Environment for Lower-Value REO Properties Table 8 shows the results of the subsample of REO properties valued at less than $250,000. Based on the distribution in Figure 10, $250,000 can be regarded as the cut- off point for the subsample, which needs further investigation. The property value of $250,000 was located below the 25th percentile in all samples. Compared with the results of all samples in Table 7, fewer variables were found to be significant in Table 8. The estimated coefficients of property and neighborhood characteristics were similar to the results of Table 7. Among property attributes, the square footage and year built were found to be significant. A hundred unit increase in the square footage decreased the likelihood of being sold by 1.81%. A one year increase in the year of construction increased the likelihood of being sold by 0.17%. From the significant variables in the socio-economic characteristics, the estimated hazard ratio of the Black population was 0.9847, meaning that the increase of 10 percentage points in the share of the Black population in a given neighborhood decreased the likelihood of an REO being sold by 1.53%.

For built environmental characteristics, residential density and street connectivity were found to be significant. Low-value REO properties in upper-level residential density had a 12.23% lower likelihood of being sold than those in middle-level residential density at the mean of the samples.26 The effect of high residential density was a decrease in the likelihood of lower-value REO properties being sold. Low-value

134

REO properties in the lower level of street connectivity had a 1.05% lower likelihood of being sold at the mean of the samples. The bike lane and park availability variables showed no differences in direction or significance.

Figure 12 shows different patterns of the plots from those in Figure 11. The plot of the upper level of residential density and street connectivity is upward, with

increasing property values; however, the estimated log hazards are lower than in the lower and middle levels. For any given levels of land-use mix, the estimated log hazard decreased with property value. The plot of the lower level of land-use mix is fairly close to a horizontal line, and the estimated log hazard is higher than in the middle and upper levels.

135

Table 8. Results of REO Duration and Built Environment for Low-Value REOs

Variables Coeff. (S.E.) Haz. Ratio (S.E.) Z P-value 2009 -0.1001 (0.0187) 0.9048 (0.0169) -5.36 0.000 2010 -0.1935 (0.0207) 0.8242 (0.0171) -9.36 0.000 2011 -0.0609 (0.0231) 0.9410 (0.0217) -2.64 0.008 2012 -0.0048 (0.0331) 0.9953 (0.0330) -0.14 0.886 2013 0.4083 (0.0848) 1.5043 (0.1276) 4.81 0.000 Sqft (/100) -0.0184 (0.0025) 0.9819 (0.0025) -7.37 0.000 Beds 0.0123 (0.0126) 1.0124 (0.0127) 0.98 0.328 Year built 0.0017 (0.0005) 1.0017 (0.0005) 3.53 0.000 Loan-to-value 0.0001 (0.0001) 1.0001 (0.0001) 0.63 0.528 Median income Middle 0.0051 (0.0209) 1.0051 (0.0210) 0.24 0.810 High -0.0371 (0.0397) 0.9637 (0.0383) -0.93 0.350 Hispanic (/10) -0.0051 (0.0059) 0.9950 (0.0058) -0.87 0.384 Black (/10) -0.0155 (0.0078) 0.9847 (0.0077) -1.98 0.048 Asian (/10) 0.0096 (0.0214) 1.0097 (0.0216) 0.45 0.653 Pop18 (/10) 0.0222 (0.0245) 1.0224 (0.0251) 0.90 0.366 Pop65 (/10) 0.0244 (0.0349) 1.0247 (0.0358) 0.70 0.485 Ownership (/10) 0.0070 (0.0080) 1.0070 (0.0081) 0.87 0.384 Unemployment (/10) 0.0125 (0.0195) 1.0126 (0.0198) 0.64 0.521 Vacancy (/10) 0.0344 (0.0181) 1.0350 (0.0188) 1.90 0.058 Mortgage (/10) -0.0041 (0.0331) 0.9961 (0.0329) -0.12 0.904 Active living (/10) -0.0008 (0.0429) 0.9993 (0.0429) -0.02 0.986 Crime density 0.0171 (0.0128) 1.0172 (0.0130) 1.34 0.180 Crash density -0.3721 (0.2265) 0.6893 (0.1561) -1.64 0.100 Property Value (PV) (/$10,000) -0.0043 (0.0033) 0.9958 (0.0033) -1.31 0.192 Residential density Lower percentile 0.0404 (0.0874) 1.0412 (0.0910) 0.46 0.644 Upper percentile -0.2627 (0.1292) 0.7691 (0.0994) -2.03 0.042 PV × Residential density Lower percentile -0.0006 (0.0047) 0.9995 (0.0047) -0.11 0.912 Upper percentile 0.0074 (0.0064) 1.0074 (0.0064) 1.16 0.245 Land-use mix Lower percentile -0.0274 (0.0712) 0.9731 (0.0693) -0.38 0.701 Upper percentile 0.0053 (0.0758) 1.0054 (0.0762) 0.07 0.944

136

Table 8. Continued

Variables Coeff. (S.E.) Haz. Ratio (S.E.) Z P-value

PV × Land-use mix Lower percentile 0.0039 (0.0040) 1.0039 (0.0040) 0.98 0.329 Upper percentile -0.0026 (0.0040) 0.9975 (0.0040) -0.65 0.515 Street connectivity Lower percentile -0.2664 (0.0888) 0.7663 (0.0680) -3.00 0.003 Upper percentile -0.1903 (0.1241) 0.8268 (0.1026) -1.53 0.125 PV × Street connectivity Lower percentile 0.0155 (0.0048) 1.0156 (0.0049) 3.23 0.001 Upper percentile 0.0061 (0.0062) 1.0061 (0.0062) 0.98 0.326 Bike lane availability 0.0449 (0.0170) 1.0459 (0.0178) 2.64 0.008 Park availability -0.0109 (0.0183) 0.9892 (0.0181) -0.60 0.551

Notes: N=20174, LR-Chi2=479.63 (p<0.0001), Log-likelihood=-175574.94; The sub-sample for lower- valued REOs (less than $250,000) was estimated; The units of Property value, Sqft, Hispanic, Black, Asian, Pop18, Pop65, Unemployment, Vacancy, Mortgage, and Active living were adjusted to obtain valid coefficients and hazard ratios; Bold texts represent the statistical significance at 0.05 level.

137

5.5 Conclusions and Policy Implications

One public policy approach for resolving foreclosure problems would be to reduce REO activity and the length of time a property remains in REO status. While “double triggers” (negative individual life events and negative home equity) may be the key to increasing mortgage default risk (Foote et al., 2008), the marketability of a property may depend more on the bundle of structural and environmental characteristics of the property. The literature has shown the importance of the environment in raising the marketability of a property. For example, walkable urban forms that provide greater accessibility place a price premium on a property. In this sense, walkable environments can help properties exit the foreclosure process by being sold to a new owner. This research utilized the D-variable (density, diversity, and design) framework, which provides measurable constructs for spatial characteristics of walkable neighborhoods, in order to establish whether and how walkable environments can help reduce REO density and duration. Although a growing number of studies argue that neighborhood context, with a particular focus on socio-economic characteristics, should be taken into

consideration in designing foreclosure policies, no study has yet examined built environmental impacts on REO density and duration.

The findings highlight that not only socioeconomic conditions but also built environmental characteristics have significant associations with REO density and REO duration. A larger percentage of blacks, unemployed, vacancies, and mortgaged homes increased the density of REOs. In REO duration, relatively few variables, such as Hispanic and black, were significant. Vacancy was also significant, but increased the

138

likelihood of the sale of an REO. Safer neighborhoods and more accessible and diverse settings of the built environment, encouraging walkability, were important

considerations in reducing REO filings. Denser and more accessible settings of the built environment increased the likelihood of a sale, but only in cases of higher-value REOs. The findings indicate that the walkable environments contribute to a strategy for increasing the marketability of properties and reduce the slide of REOs into deterioration.

In the REO duration analyses, this research revealed further implications through testing the interaction terms between assessed property values and built environmental attributes. Holding built environmental factors constant at the middle level, this study found that low-value REOs tended to be sold more quickly, which is a similar finding to previous research on the relationship between market value and REO duration. However, in neighborhoods with a high density of residents and street connections, low-value REOs were less likely to be sold, but high-value REOs usually sold faster. This finding may reflect the fact that denser and more accessible neighborhoods translate inequitably into the marketability of properties. In addition, foreclosure disparities may exist in an inequitable distribution of market-supported environments. The samples used in this research also showed that the geographic distribution of lower-valued REO properties was more concentrated in low-income and minority communities. It is possible that the quality of built environmental resources is less desirable in such communities, and that the resources are less available, even though those communities have the same spatial structures of built environmental attributes as high-income communities. The evidence

139

has shown that low-income and minority communities are generally disadvantaged in neighborhood safety such as crimes and crashes, and in environmental features such as esthetics and recreational areas and facilities (Sallis et al., 2011; Zhu & Lee, 2008). This research emphasizes the gap in policy intervention, which requires further attention on built environmental attributes.

The findings of this research are constrained by the following limitations. First, due to the data availability, this research did not include profiles of sellers and brokers. Because of the seller’s motivation, an REO property might be held off as “shadow inventory” until the market condition recovers or “dumped” because of a difficult market condition. A broker’s ability may also influence the REO sales (Y. Li & Walter, 2013). Most covariates (such as property attributes and neighborhood characteristics) used in this research were estimated consistently with the previous literature, but the omitted variables would merit further investigation in future research. Second, this research focused on REO properties, but the data did not divulge information on whether REO properties are vacant or tenant-occupied. Third, due to the geographical (LA County) and temporal (2008-2013) limitations, the findings may not apply to other contexts and other time periods. REO accumulation may vary across states where the foreclosure processes are different (Immergluck, 2010a). Last, this research did not include specific aspects of built environmental resources, such as the availability and quality of

neighborhood amenities. This would be the potential for uncovering further relationship between built environments and foreclosures in future research.

140

To help local communities recover from the foreclosure crisis, policymakers and local governments focus on the effectiveness of existing policy interventions such as the Neighborhood Stabilization Program (NSP), to reduce vacancies and rehabilitate

communities by encouraging the purchase of foreclosed homes. As strategic options for designing effective policies, further enforcement efforts for improving the quality of environmental attributes are also needed to help protect our neighborhoods from the impacts of foreclosures.

141

CHAPTER VI CONCLUSION

This dissertation has examined how walkable environments alleviate foreclosure- related activities. This study advances the existing literature by incorporating built environmental factors into the examination of foreclosure spillover effects, foreclosure density, and foreclosure duration. Findings from this study provide important

implications for future interventions to reduce the impacts of foreclosure.

6.1 Overview of Findings

Chapter III presented a comprehensive examination of the literature to identify methodological and content issues and improve understanding of foreclosure spillover effects on property values. This review highlighted a lack of a thorough examination of contextual neighborhood factors, such as built environmental effects. Previous evidence supports the economically sustainable benefits of the walkable environment as a result of accessible and compact urban design. Thus, the review discussed opportunities for future research that addresses environmental interventions to reduce the price spillover effects of foreclosures. The review also indicated that insufficient attention has been given to the dynamic and elaborate details of foreclosure measurement. Measuring specific foreclosure stages and statuses, property types, and property conditions may provide opportunities to disentangle the mechanisms affecting foreclosures and nearby property values. In addition, further research is suggested to examine the extent to which

142

foreclosure spillover effects vary across neighborhood characteristics, housing market periods, and housing submarkets.

The research gaps addressed in Chapter III and Chapter IV examined how neighborhood walkability can weaken the negative foreclosure spillover effects on property values. By using the interaction terms between neighborhood walkability, measured as Walk Score (WS) and neighboring foreclosures, this chapter evaluated the degree to which neighborhood walkability lessens the intensity of foreclosure spillover effects on property values. Using separate models, the differential impacts of the walkability premium on price spillovers of foreclosures were also analyzed for two different income groups (low versus high) and market periods (the housing market crash of 2010 versus the housing market recovery of 2013). The results showed that the price spillover effects of foreclosure were significantly attenuated in very walkable

neighborhoods (WS: 70-95) by 54.26-69.77% for 2010, and 66.04-84.91% for 2013; however, the mitigation effects were insignificant for low-income groups. This leads to the conclusion that potential income disparities in walkability impacts may exist, which ameliorate negative price spillovers of foreclosures.

Chapter V investigated the influence of walkability-related environments – residential density, land-use mix, and street connectivity – as represented by the D- variable frame (density, diversity and design), on real estate owned (REO) foreclosures. Two dependent variables were used: REO density and REO duration. This chapter highlighted a lower REO density in neighborhoods which are safer, more accessible, and have a diverse setting of built environments. By using interaction terms between the

143

market value of REOs and built environmental factors, the study found that higher-value REOs in denser and more accessible neighborhoods were more likely to be sold; on the contrary, lower-value REOs were less likely to be sold. These findings help to explain inconsistent results regarding the duration of low-value REOs in the previous literature. This study implies the underlying disparity issues regarding environmental support for the marketability of foreclosed properties. Essentially, compact and accessible

environments are not beneficial to low-income and minority communities.

Overall, this dissertation demonstrates that a walkable neighborhood has the potential to provide resilience benefits of enhancing neighborhood stability in the aftermath of an economic shock. In particular, this dissertation suggests that

neighborhood walkability can help to boost recovery from a foreclosure crisis through producing economic benefits and reducing foreclosure-related events. As emphasized by Jane Jacobs (1961), the findings from this dissertation imply that more accessible and compact urban designs could help holistic planning strategies to enhance social and